MPPT Solar Charging

The problem: a panel only gives full power at one operating point

A solar panel is not a battery. A battery holds its terminal voltage roughly fixed and delivers whatever current you draw. A panel does the opposite: its voltage and current are locked together by a curve, and where you sit on that curve is decided by the load you connect — not by the panel. Connect a short circuit and the panel delivers its full short-circuit current I_sc at zero volts, so zero power. Connect nothing and it sits at its open-circuit voltage V_oc at zero current, again zero power. Somewhere between those two extremes is a single operating point where the product of voltage and current — the power — is maximized. That point is the maximum power point (MPP), and it sits at the knee of the I-V curve.

The whole job of MPPT solar charging is to keep the panel parked at that knee while the sun, the clouds, and the panel temperature all conspire to move it. The reason this is non-trivial is that the battery you are charging has its own opinion about voltage. A 12 V lead-acid battery wants to be charged at about 13.5-14.4 V. A 36-cell "12 V" panel has its MPP at roughly 17-18 V. If you simply wire the panel to the battery — which is what the cheapest controllers do — the battery clamps the panel down to 14 V, dragging it off the knee and onto the steep part of the curve where the same current is delivered at a lower voltage. The power above 14 V is simply discarded.

The I-V curve and the power curve

A single-diode model captures the panel's behaviour well enough to reason about it. The terminal current is the light-generated current minus the diode recombination current:

I = I_ph − I_0 · ( exp( (V + I·R_s) / (n·V_T) ) − 1 ) − (V + I·R_s) / R_sh

where I_ph is the photocurrent (proportional to irradiance), I₀ the diode saturation current, V_T = kT/q the thermal voltage (~25.7 mV at 25 °C), n the ideality factor, and R_s, R_sh the series and shunt resistances. You do not solve this on a microcontroller in real time — you just need its shape. For most of the curve the panel behaves like a current source (flat top, current ≈ I_sc), then near V_oc the exponential diode term takes over and current collapses. Multiply current by voltage at each point and you get the power curve: a single smooth hump with one peak. That single-peak shape is exactly what makes hill-climbing trackers work.

The peak sits at the maximum power point: P_mp = V_mp · I_mp. Two ratios characterise a panel here. The fill factor FF = (V_mp·I_mp) / (V_oc·I_sc) is how "square" the curve is — 0.7-0.8 for good crystalline silicon. And as rules of thumb, V_mp ≈ 0.80-0.82 · V_oc and I_mp ≈ 0.90-0.95 · I_sc. For a typical 60-cell module: V_oc ≈ 38 V, I_sc ≈ 9 A, V_mp ≈ 31 V, I_mp ≈ 8.5 A, so P_mp ≈ 265 W with a fill factor around 0.77.

Why the maximum power point is a moving target

If the MPP stood still, you would calibrate it once and be done. It does not. Two effects move it continuously:

• Irradiance scales the current. The photocurrent I_ph is almost exactly proportional to the sunlight hitting the panel. Halve the irradiance and the whole I-V curve drops vertically — I_sc and I_mp roughly halve while V_oc moves only slightly (it depends logarithmically on current). A cloud edge can cut the MPP current in half in well under a second, so the tracker must respond fast.
• Temperature shifts the voltage. The open-circuit voltage of crystalline silicon falls by about −2.2 mV per cell per °C, which is roughly −0.30 to −0.35% of V_oc per °C. A 60-cell panel that sits at V_mp = 33 V on a 0 °C morning may be at 27 V when the cells reach 65 °C in full summer sun. That is why cold-and-bright is the best case for solar harvest and the worst case for a PWM controller: the panel is at its highest voltage exactly when PWM is throwing away the most.

So the MPP wanders across the voltage axis with temperature and up and down the current axis with irradiance, sometimes within the same minute. The tracker's job is to ride the moving peak, not just find it once.

The converter underneath: how MPPT decouples panel from battery

The thing that lets the panel sit at 31 V while the battery sits at 14 V is a switching DC-DC converter — usually a buck (step-down) converter, sometimes a boost or buck-boost depending on the voltage relationship. The converter is the actuator; the MPPT algorithm is the controller. The key relationship for an ideal buck converter is that the output voltage is the input voltage times the switch duty cycle:

V_out = D · V_in          (ideal buck, D = on-time / period)
P_in  = P_out / η         (η ≈ 0.96 to 0.99)
so:    I_out = η · (V_in / V_out) · I_in

The duty cycle D is the single knob the algorithm turns. Changing D changes the effective resistance the panel sees, and therefore where on the I-V curve the panel operates. Push D one way and the panel voltage rises toward V_oc; push it the other way and it falls toward the current-source region. The algorithm watches panel power and adjusts D until power peaks. Once at the MPP, the converter transfers that power to the battery: because output voltage is lower than input, output current is correspondingly higher — this is the "current boost" MPPT users talk about. A panel delivering 8.5 A at 31 V (265 W) into a 14 V battery puts out roughly 31/14 × 8.5 × 0.97 ≈ 18.3 A, far more than the 8.5 A a PWM controller would pass.

Perturb-and-observe: hill-climbing in a few lines of code

The dominant algorithm is perturb-and-observe (P&O). It needs only a panel-voltage measurement and a panel-current measurement, and it carries one bit of memory — which direction it last moved. Each control cycle:

measure V, I            // panel side
P = V * I
dP = P - P_prev
if (dP > 0)             // power rose — keep going the same way
    D += step * dir
else                    // power fell — turn around
    dir = -dir
    D += step * dir
P_prev = P

The controller climbs the power hump and then, having nowhere higher to go, oscillates by one step on either side of the peak forever. That oscillation is the algorithm's signature cost: it spends part of its time slightly off-peak. The step size sets the tradeoff. A large step tracks a fast-moving cloud edge quickly but wastes power oscillating widely around the peak; a small step sits tight on the peak in steady state but lags badly when the light changes. Modern controllers use a variable step — large when far from the peak, shrinking as dP approaches zero — to get both fast acquisition and tight steady-state holding.

Incremental conductance is the standard refinement. At the MPP, dP/dV = 0, which expands to d(VI)/dV = I + V·dI/dV = 0, i.e. dI/dV = −I/V. The algorithm measures the incremental conductance dI/dV and compares it to the instantaneous conductance −I/V: if they are equal it is exactly at the peak and stops perturbing; if dI/dV > −I/V it is left of the peak and moves right; otherwise it moves left. This lets the controller actually stop at the MPP rather than dither around it, cutting steady-state loss. Constant-voltage is the crude fallback: hold the panel at a fixed fraction (~0.76) of V_oc, which is cheap but ignores temperature drift and irradiance-dependent curvature.

Worked example: cold morning vs hot noon

Take a 60-cell, 300 W-class module charging a 24 V (nominal ~28 V charging) battery bank through a buck MPPT controller. Datasheet at standard test conditions (25 °C cell): V_mp = 31.5 V, I_mp = 9.5 A, P_mp = 299 W, temperature coefficient of V_oc = −0.31%/°C.

Cold bright morning, cell temp = 5 °C  (20 °C below STC):
  V_mp shift  = +0.31%/°C × 20 °C × V_mp ≈ +1.95 V  →  V_mp ≈ 33.5 V
  I_mp        ≈ 9.5 A (full sun)
  P_panel     ≈ 33.5 × 9.5 ≈ 318 W

  PWM controller: panel forced to ~28 V battery
    P_PWM     ≈ 28 V × 9.5 A ≈ 266 W      (off the knee, low side)
  MPPT controller: panel at 33.5 V, η = 0.97
    P_MPPT    ≈ 318 × 0.97 ≈ 308 W
  MPPT gain   ≈ 308 / 266 − 1 ≈ +16%   (and more vs a 12 V battery)

Hot noon, cell temp = 65 °C  (40 °C above STC):
  V_mp shift  = −0.31%/°C × 40 °C × V_mp ≈ −3.9 V  →  V_mp ≈ 27.6 V
  I_mp        ≈ 9.5 A
  P_panel     ≈ 27.6 × 9.5 ≈ 262 W
  PWM:  28 V battery is now ABOVE V_mp → panel barely delivers
  MPPT: holds 27.6 V, P_MPPT ≈ 262 × 0.97 ≈ 254 W

The cold case shows the headline 15-20% gain; the hot case shows something subtler and more important. When the battery voltage rises above the panel's V_mp — common with hot panels and a high battery state of charge — a PWM controller cannot even reach the MPP because it can only pull the panel down to battery voltage, never operate it independently. The MPPT converter holds the panel at 27.6 V regardless of battery voltage, which is exactly the case where the dumb controller collapses. This is also why high-voltage panel strings into low-voltage batteries are the MPPT sweet spot: the bigger the V_mp-to-battery gap, the more there is to recover.

MPPT vs PWM: where the difference actually lands

Property	MPPT controller	PWM controller
How it connects	DC-DC converter (buck/boost) between panel and battery	Switch that ties panel directly to battery
Panel operating voltage	Held at true Vmp (e.g. 31 V)	Dragged to battery voltage (e.g. 14 V)
Energy harvest	15-30% more, climate-dependent	Baseline (loses everything above battery V)
Panel : battery voltage	Any — can step a 150 V string into a 12 V battery	Must be matched (panel Vmp close to battery V)
Best conditions	Cold, bright, high voltage gap	Warm, panel Vmp ≈ battery V
Wiring	Thinner wire OK — low current at high string voltage	Full panel current on the run — thicker wire
Internal loss	1-4% switching + conduction	<1% (just a switch)
Cost / complexity	Higher — inductor, switch, sense, MCU	Lower — switch + simple regulator
Partial-shade handling	Can run a global sweep to find the true peak	None — sits at battery voltage regardless

The honest summary: PWM wins on cost and is fine for a small, warm-climate system where the panel voltage already matches the battery. MPPT wins everywhere the voltages diverge, the panels run cold, or the wire runs long — which is most serious installations.

Partial shading and the multiple-peak trap

Everything above assumes a single smooth power hump. Partial shading breaks that assumption. Cells in a panel are wired in series, so a single shaded cell would otherwise drag the whole string's current down to its own reduced level. To prevent that — and to stop the shaded cell from being reverse-biased into a hot spot — panels include bypass diodes across each sub-string (typically three per 60-cell module). When a sub-string is shaded, its bypass diode conducts and the current routes around it. The price is that the I-V curve develops a step, and the power curve grows multiple local maxima.

A vanilla perturb-and-observe tracker climbs to the nearest peak and stops — which may be a local maximum far below the true global peak, costing a large fraction of the available power. The fixes are well established:

• Periodic global sweep. Every few minutes the controller sweeps the panel voltage across its whole range, records power at each point, then jumps back to the highest peak found. Cheap in software, costs a brief harvest dip during the scan.
• Module-level power electronics (MLPE). Put an MPPT on each panel — micro-inverters (Enphase) or DC power optimizers (SolarEdge, Tigo). Each module tracks its own peak, so one shaded panel cannot pull a whole string off its peak. Standard on rooftop residential, where chimneys and trees guarantee shading.
• String layout. Group panels likely to be shaded together so the loss is contained, rather than scattering shade across many strings.

Failure modes and trade-offs in practice

• Step-size mistuning. Too large and the tracker oscillates wide of the peak, visibly losing power and stressing the output; too small and it loses the peak whenever a cloud moves. Variable step size or adaptive P&O is the cure.
• Tracking confusion under fast clouds. If irradiance changes between the two measurements P&O uses to decide direction, the algorithm can misread a power change caused by the cloud as a result of its own perturbation, and walk the wrong way. Robust implementations measure faster than the light changes or use irradiance-compensated P&O.
• Local-maximum lock under shade. Covered above — needs a global search or per-module MPPT.
• Converter thermal limits. The buck inductor and switching MOSFET dissipate the conduction and switching losses; on a 100 A controller that can be tens of watts, so heatsinking and current derating at high ambient temperature matter. Many controllers fold back output current above ~45 °C heatsink temperature.
• Input over-voltage. A cold morning raises V_oc; a long series string of cold panels can exceed the controller's maximum input voltage and destroy the input stage. String sizing must use the coldest-expected-temperature V_oc, not the STC value.
• Battery-stage interaction. MPPT maximizes harvest, but a full battery cannot absorb the power. The controller must hand off from MPP tracking to a constant-voltage / constant-current charge profile (bulk, absorption, float) as the battery fills, deliberately moving the panel off the MPP to curtail current. Confusing "maximize harvest" with "always run at MPP" overcharges batteries.

Where MPPT shows up

• Standalone solar charge controllers. Victron SmartSolar, Morningstar, EPEVER — buck MPPT converters from 10 A to 100 A feeding 12/24/48 V battery banks in RVs, boats, and off-grid cabins.
• Grid-tie string inverters. The front end of every string inverter (SMA, Fronius, Huawei) is an MPPT stage — often two or three independent trackers so different roof faces with different sun angles each find their own peak.
• Module-level electronics. Enphase micro-inverters and SolarEdge optimizers run one MPPT per panel for shade tolerance and per-module monitoring.
• Spacecraft power. Satellite solar arrays use MPPT (or the related direct-energy-transfer schemes) because array temperature swings hundreds of degrees between sun and eclipse, moving the MPP far.
• Solar water pumping and small PV-direct loads. A dedicated MPPT lets a pump motor draw the panel's full power across changing sun without a battery buffer.